Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
CITATIONS
READS
57
199
3 authors:
Espen Strand
Christian Jrgensen
University of Bergen
SEE PROFILE
SEE PROFILE
Geir Huse
Institute of Marine Research in Norway
76 PUBLICATIONS 3,148 CITATIONS
SEE PROFILE
University of Bergen, Department of Biology, P.O. Box 7800, N-5020 Bergen, Norway
b Institute of Marine Research, P.O. Box 1870, Nordnes, N-5817 Bergen, Norway
Received 18 June 2004; received in revised form 13 December 2004; accepted 17 December 2004
Abstract
We present a bioenergetic model for buoyancy regulation that incorporates the restrictions and costs of swimbladder regulation
with four means of hydrodynamic lift production: hovering, swimming with extended pectoral fins, swimming with adjusted tilt
angle, and body lift. Previous models addressing vertical migration in fish with swimbladders have either assumed no energetic
cost or a static cost of vertical migration. In this model, parameterised for Atlantic cod Gadus morhua, existing theory and
experimental data on bioenergetics, physiology, and hydromechanics are integrated. Trade-offs and limitations were investigated
from a behavioural perspective. Regulation of swimbladder volume is energetically cheap but slow. Because of the asymmetry in
absorption and secretion rates with depth, fish that perform vertical migrations regularly will often be negatively buoyant and will
therefore experience additional energy costs associated with hydrodynamic lift production. Hovering was optimal for slightly
negatively buoyant fish, whereas tilted compensatory swimming was optimal in all other situations. The predicted optimal tilt
angle was approximately 7 , and increased for small fish and for fish that were close to neutrally buoyant. The energetic saving
from tilting was small for cod, and potential conflicts with other behaviours may determine when tilting would be preferential in
nature. Swimbladder volumes and the corresponding energetic costs were calculated for constructed vertical migration patterns
and on a set of depth data from a free-living cod. For the free-living cod, the model predicted that the fish would be neutrally or
negatively buoyant, with the swimbladder being up to 40% smaller than the optimal volume.
2005 Elsevier B.V. All rights reserved.
Keywords: Atlantic cod (Gadus morhua); Bioenergetics model; Buoyancy regulation; Physoclist swimbladder; Tilt angle; Vertical migration;
Target strength
1. Introduction
0304-3800/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolmodel.2004.12.013
Living tissue is, with the exception of lipids, heavier than water. Aquatic organisms that intend to utilise
the free water masses will therefore have to adopt
some mechanism by which to control buoyancy. Nature
310
Buoyancy regulation has hitherto not been mechanistically quantified in behavioural models studying the
trade-offs involved in vertical migration. Earlier models have either assumed the cost of vertical migrations
proportional to depth change (e.g. Rosland and Giske,
1994; Strand et al., 2002), assumed no cost (e.g. Clark
and Levy, 1988), and have furthermore not identified
the constraints imposed by a swimbladder. To improve
the ecological realism of such models, this work pieces
together knowledge from three research areas: swimbladder physiology, hydrodynamics of lift, and bioenergetics.
Comparisons between species have sketched the
broad perspective of advantages and disadvantages
with the different means of buoyancy regulation.
Changes in swimbladder volume are constrained by
time and depth, and hydrodynamic lift by the energy
requirements. The model presented here describes the
processes involved in buoyancy control: swimbladder
volume changes, hovering, swimming with fins as hydrofoils and with a tilt angle. These different modes
of buoyancy regulation have different rates, different
energetic costs, and different constraints, and have
been well-studied separately (Alexander, 1971; Blake,
1979; Harden Jones and Scholes, 1985; Vogel, 1994).
By formulating them in a common bioenergetics framework, the trade-offs between the swimbladder and hydrodynamic forces as means of buoyancy regulation
can be compared. The model can answer two lines of
questions: (1) on the shorter time-scale, which types of
behaviour are possible with regard to the physiological
constraints involved, and (2) in the longer run, what
are the energetic costs of the different modes of buoyancy regulation. Insights from this last approach can
be used to interpret field data from free-ranging fish,
and the bioenergetics currency allows comparisons of
costs and benefits of different foraging behaviours to be
studied. By implementing the model presented here as
a submodel in behavioural models, questions into the
adaptive nature of foraging behaviour and other types
of behaviour involving vertical migrations in pelagic
fish can be investigated.
Fish have adapted to a great range of environments,
and the constraints that a swimbladder forces upon
the general physoclist fish will be hard to examine.
To better investigate the models properties, speciesspecific values for Atlantic cod (Gadus morhua) were
used. This is the physoclist species that has been most
2. Model description
The equations that describe buoyancy control in this
model are intended for recalculation over several timesteps. For every time-step, the following procedure
takes place: (1) the fish depth is used as input; (2) leakage from the swimbladder is calculated based on this
depth; (3) the buoyancy experienced by the fish at the
actual depth is estimated based on the new state of its
swimbladder (change of depth and leakage); (4) maximum rates for secretion and absorption of gases are calculated. The required secretion/absorption takes place
within these limits; (5) the buoyancy not accounted for
by changes in swimbladder volume has to be generated
by hydrodynamic forces, i.e. through increased swimming at the optimal tilt angle, or by hovering; (6) the
metabolic costs of secretion and compensatory swimming are summarised.
2.1. The swimbladder
Physoclist fish have the ability to secrete gas through
a gas gland on the swimbladder wall (Prosser, 1973).
The gas gland produces lactic acid and CO2 , which
lower the pH and increase the concentration of solutes
in the blood. Direct and indirect effects effectuate the
release of haemoglobin-bound oxygen and physically
dissolved gases through the Bohr, Root, and saltingout effects (Jobling, 1995; Schmidt-Nielsen, 1997; Pelster, 2001). In addition, the arterial blood supply flows
through a counter-current vascular arrangement (rete
mirabile) that multiplies blood gas tensions and maintains pH and solute concentrations in the gas gland.
As a result, gas tension in the gas gland may exceed
311
that of arterial blood several-fold, and gases thus diffuse into the pressurised swimbladder (Scholander and
van Dam, 1954; Kuhn et al., 1963; Steen, 1963a).
During ascents, physoclist fish will reabsorb the excess gas inside the swimbladder through a particularly
vascularised area called the oval (Fange, 1953; Steen,
1963b). The rate of absorption is limited, however, and
rapid ascents can thus be dangerous to physoclist fish,
as the swimbladder may expand beyond control or burst
(Harden Jones, 1952; Tytler and Blaxter, 1973).
In this model, leakage and the maximum sustainable rates of secretion and absorption are based on gas
composition with emphasis on oxygen, while the energetic cost of secretion is calculated based on physical work and a biological efficiency factor. Wherever
physically dissolved gases are involved, all gases in
the swimbladder are assumed to have the same physical
properties as oxygen. This approximation can be partly
justified since the next two components, CO2 and N2 ,
have a higher and lower solubility, respectively. Furthermore, other gases (Ar, Ne, He) can be found in the
swimbladder to varying degrees, and the gas composition will hence change in a complex manner resulting
from past vertical behaviour. The equations relating to
haemoglobin-bound oxygen use a partial pressure of
oxygen in the swimbladder that is a constant fraction
of the hydrostatic pressure. Because CO2 has a higher
solubility than O2 , this will tend to underestimate absorption rates during the initiation of absorption (when
CO2 would have been removed faster), but will slightly
overestimate rates after the majority of CO2 has been
removed (which means that N2 would have been more
important for the overall rate of absorption) (Steen,
1963b and references therein). The model can be elaborated to incorporate gases with different physical properties, but at this point such a level of detail was not
found worthwhile. The partial pressure of oxygen and
other gases would then become dynamic, and the model
could thus also shed light on how behavioural decisions
influence the actual gas composition of the swimbladder. Although a full modelling of all gas components
may be possible, it is beyond the scope of this paper.
SI units have been preferred throughout this paper,
although it may seem awkward to measure the volume
of the swimbladder of a small cod in cubic metres. This
makes the equations easier to read, and will thus hopefully enhance understanding of the concepts involved.
See Table 1 for details on all variables.
312
Table 1
Variables and parameters used in the model for buoyancy regulation
Description
Value
Unit
0.050.20
Degrees
mol s1
N
m
m3 s1
J s1
m
N
mol s1
kg
J s1
J s1
mol s1
J s1
P
Ss
Thrust
U; UH
Umax
Vn ; V s
z; zc
Tilt angle
Energy conversion efficiency during swimming
Maximum rate of O2 absorption
Increase in energy expenditure during swimming
Buoyancy force
Body length
Cardiac output
Energy used for gas secretion
Pectoral fin length (at right angle from body)
Lift and drag force, subscript referring to fin, body and tilt
angle ()
Oxygen leakage from swimbladder
Body mass
Metabolic cost of hovering
Additional metabolic cost of extended fins
Maximum O2 that can be secreted into swimbladder; actual
O2 secreted
Standard metabolic rate; total metabolic rate including fin
drag; active metabolic rate (measured metabolism-SMR)
Hydrostatic pressure
Swimbladder surface area
Power produced during swimming
Fish swimming speed; horizontal component
Maximum sustainable swimming speed
Neutral swimbladder volume; current swimbladder volume
Depth; critical depth
1 + (z/10)
atm
m2
N
m s1
BL s1
m3
m
Parameters
;
O2
f ; w
0.397; 0.828
0.63
1081; 1026
Radians
kg m3
Hb
a/b
Crete ; Coval
Eeff
F (T)
G
g
HGROUND
[L/D]body
O2a
O2sol
P0
R
0.15
10
0.10; 0.25
0.10
0.502 at 5 C
1.5 109
9.81
1
0.75
0.10
0.04
1
8.206 105
8.3144
m3 O2 m2 atm1 s1
m s2
m3 O2 m3 blood1 atm1
atm
m3 atm mol1 K1
J mol1 K1
Reff
T
Rete efficiency
Ambient temperature
0.20
278.15
Variables
Absmax
Act
B
BL
Cout
Esec
FL
L; D
Leakage
M
MRHOVER
MRPECT
Secmax ; Sec
SMR; TMR; AMR
Species-specific values refer to Atlantic cod (Gadus morhua). Variables and parameters without dimension are denoted in the unit column.
A submerged fish has neutral buoyancy when it displaces a volume of water with a mass equal to its body
mass (Archimedes principle). The volume of a swimbladder giving neutral buoyancy (Vn ; m3 ) to a fish with
body mass M (kg) can then be calculated given the density of the fish tissue (f ; kg m3 ) and the density of
water (w ; kg m3 ):
Vn =
M(1 (w /f ))
w
(1)
both w and f can be modelled as variables to reflect heterogenic environments and ontogenetic and
seasonal changes for the particular species (discussed
in e.g. Ona, 1990).
The amount of gas (mol) in the swimbladder was
calculated using Boyles law (Harden Jones, 1951;
Alexander, 1959). If the actual swimbladder volume
(Vs ; m3 ) differs from Vn , the fish will experience positive or negative buoyancy force (B; N):
B = (Vs Vn )w g
313
(2)
Leakage =
GSs (P P0 )
RT
(4)
where G (m3 O2 m2 atm1 s1 ) is the oxygen conductance of the swimbladder wall, R (m3 atm mol1 K1 )
the universal gas constant, and T (K) the temperature
in Kelvin.
2.4. Secretion
The equations for gas secretion are based on work by
Harden Jones and Scholes (1985). It is assumed that the
diffusion barrier between the gas gland and the swimbladder is so little that the rate-limiting step is the transport of oxygen and multiplication of partial pressures
in the rete and the gas gland. As a consequence, secretion will be independent of depth. The model calculates
the maximum rate of secretion (Secmax ; mol s1 ), as a
function of cardiac output (Cout ; m3 s1 ), the fraction of
Cout diverted to the rete (Crete ; dimensionless) (Pelster
and Scheid, 1992), the volume fraction of O2 in arterial
blood (O2a ; dimensionless) measured at P0 (1 atm), and
an efficiency factor of the rete (Reff ; dimensionless):
Secmax =
(5)
314
It is further necessary to divide the minimum energy with an efficiency factor for converting biological
chemical energy into mechanical energy (Eeff , dimensionless). Eeff includes the effects and energetic costs
for the Bohr, Root, and salting-out effects, production
of lactate, and the pentose-phosphate pathway (Pelster
et al., 1989, 1994). If these effects were to be modelled explicitly, it would be necessary to know the cost
and contribution of each of these effects. A precise set
of equations to describe these processes thus requires
further experimental research; in this model the focus
is set on the overall energetic cost of secretion. Esec
(J s1 ) is thus the required energy to secrete the gases:
P
1
Esec = Sec RT ln
(7)
Eeff
P0
where P is the hydrostatic pressure and P0 the combined partial pressures of the gases in arterial blood,
assumed constant (1 atm). Sec is the amount of gases
secreted and always less than or equal to Secmax . The
depth at which the leakage rate equals the maximum
gas secretion rate is termed the critical depth (zc ; m),
and was calculated by finding the pressure at which
leakage from Eq. (4) equals maximum secretion from
Eq. (5) (P = l + z/10):
Secmax RT
(8)
zc = 10
GSs
2.5. Absorption
Physoclist fish have a particular area on the swimbladder wall called the oval, which is specialised for
gas absorption (Fange, 1953; Steen, 1963b). There are
three major variables that influence the rate of oxygen
absorption (Steen, 1963b). First, only the fraction of
the cardiac output that flows through the oval (Coval ;
dimensionless) can absorb oxygen. Second, the pressure gradient between the swimbladder and the arterial
blood (in this model calculated as P P0 ) determines
the amount of gas that can be physically dissolved.
Third, available oxygen-binding sites on haemoglobin
can transport additional oxygen. By adopting the
method of Harden Jones and Scholes (1985), the maximum oxygen absorption rates (Absmax ; mol s1 ) can
be calculated as:
Cout Coval P0 O2sol (P P0 )
Absmax =
+ O2a Hb
RT
O2
(9)
(10a)
(J s1 )
is
(10b)
315
Fig. 1. Schematic drawing of the forces acting on a fish with swimbladder. (a) The forces acting on a negatively buoyant fish swimming
horizontally. Thrust will equal drag at constant velocity. Lift is created by using the pectoral fins as hydrofoils (Lfin ) and by the general
body surface (Lbody ). (b) When the fish is tilting, the thrust vector
can also produce lift (L ). This will also affect the efficiency with
which the pectoral fins and the body can create lift.
(12)
AMR
U
(13)
where min and max are minimum and maximum values for , respectively, and Umax is the maximum sustained swimming speed measured in BL s1 . Typically,
would scale between min = 0.05 and max = 0.20.
Defining in this way preserves Dbody as a function of U22.5 , and lets scale with body size in accordance with the biological interpretation. For most
species Umax decreases with body size and increases
with temperature.
Given that the fish maintains a constant speed, the
horizontal component of the thrust vector will be equal
to the body drag Dbody (N):
Dbody = cos Thrust
(15)
316
(Lfin / cos )2
w UH FL2
(17)
(18)
to width ratio (the a to b ratio in Eq. (3)) was estimated to be 10. Scholander (1956) found that O2 made
up on average 63% of the secreted gases in cod, thus
O2 = 0.63. The energy converting efficiency during
secretion was set to Eeff = 0.10 (Alexander, 1971). The
oxygen content of arterial fish blood was set to 10% by
volume (O2a = 0.10, Prosser, 1973; Harden Jones and
Scholes, 1985). Steen (1963a) obtained experimentally
a rete efficiency of 20% in the eel (Anguilla anguilla),
thus Reff = 0.20. These parameters describe secretion in
accordance with the experiments in Harden Jones and
Scholes (1985).
The total metabolic cost (TMR; Eq. (10b)) was
calculated based on a general bioenergetic model
by Hewett and Johnson (1992) with cod parameters
(Hansson et al., 1996):
SMR = M F (T )
(19a)
(19b)
Act =
3.2U 1.5
BL
(19c)
where M is a weight dependent and F(T) a temperature dependent function of metabolism with = 0.397,
= 0.828, and F(T) = 0.502 at 5 C (see Hansson et al.,
1996 for details). Act (dimensionless) is the incremental cost of activity and is estimated based on measurements of oxygen consumption during swimming (U)
for cod in the range 0.241.91 kg (expressed on the
form used in Ware, 1978; Schurmann and Steffensen,
1997; Webber et al., 1998; Reidy et al., 2000).
Cardiac output (Cout , m3 s1 ) based on data from
Webber et al. (1998) was calculated as a function of
body mass and the metabolic rate corresponding to a
cruising speed of 0.4 BL s1 :
Cout = 9.07 107 TMR 3.79 108 M
(20)
(21)
317
(22)
surface and body shape, and resulted in a slight asymmetry in the energetic cost between creating positive
and negative lift.
To our knowledge, no study has ever combined measures of different aspects of buoyancy regulation to a
degree that can be used as a set of independent data
to validate the model described above. That would require simultaneous measures of gas rates, buoyancy,
swimming speed and mode, and metabolic rate. Rather,
318
swimbladder dynamics and hydrodynamic lift generation has to be tested separately against experimental
datasets, and the combination of the two in a bioenergetic framework can at present only be shown, not
tested. To compensate this, we have tried to use parameter values from published experiments and strived
to make our thoughts and assumptions as clear to the
reader as possible. We have also taken two approaches
to demonstrating the benefits from combining swimbladder dynamics and hydrodynamic lift generation.
First, we show estimates for the energetic cost of buoyancy control for a series of constructed patterns for diel
vertical migration for a 2-kg cod. Each migration pattern was simulated for 6 days to stabilize, and results
were shown for the last 24-h period only. Second, we
visualise the model using 12 months of data on vertical positions collected from a 3-kg cod carrying a data
storage tag (resolution 2 h, interpolated to time-steps of
15 min, from God and Michalsen, 2000). The cod was
tagged on the 21 March 1996. To assure that any possible handling wounds would have healed, three weeks,
starting at 22 October 1996, were selected for visualisation.
3. Results
Several authors have studied different modes of
buoyancy regulation (e.g. Steen, 1963a; Alexander,
1972; Harden Jones and Scholes, 1985; Vogel, 1994).
The results presented here provide the basics of the
different modes, but focus on the link between swimbladder dynamics and hydrodynamics made possible
in our paper by the common bioenergetics approach. It
is in the linking of the different modes of buoyancy regulation in a common currency that our work transcends
earlier work and creates a basis for future behavioural
models and field data interpretation.
3.1. Swimbladder dynamics
The swimbladder is subject to the physical laws for
gases, which make swimbladder dynamics dependent
on both depth and body size. To clarify the picture, we
will describe how rates, in mol s1 , vary with depth,
and then focus on how this is translated into volume
changes and affect the behavioural repertoire.
319
Fig. 4. Properties of hydrodynamic lift production when the swimbladder volume deviates from optimal. The left-to-right axis is the
swimbladder volume, where 0% corresponds to the optimal volume
that would have resulted in neutral buoyancy. Negative buoyancy
(sinking) is to the left; positive buoyancy (floating) to the right. Fish
size (kg) increases from 0.2 kg in front to 5.0 kg in the back. (a) Compensatory swimming speed, expressed in body lengths per second
(BL s1 ). (b) The values taken by the energy conversion efficiency
between the fish oxygen consumption and the physical work required to overcome drag. increases with increasing swimming
speeds up to maximum sustainable swimming speed, where it would
reach a maximum of 0.20. (c) The optimal tilt angle at different levels
of buoyancy and fish size. Panels (a) and (b) assume that the fish was
swimming at the optimal tilt angle.
expected concave-up shape: the deviation in swimbladder volume is proportional to the required lift, which is
proportional to drag, which again is a function of U23 .
This was achieved by the energy conversion efficiency
320
321
322
Fig. 7. (a) The observed vertical position of one Atlantic cod over a period of three weeks (data from God and Michalsen, 2000). (b) The models
prediction of the cods experienced swimbladder volume expressed in percent deviation from the optimal swimbladder volume that would have
given neutral buoyancy. The cod was assumed to be neutral at time 0 and the first day is not shown to allow the swimbladder volume to stabilise.
The horizontal line is optimal swimbladder volume. (c) The corresponding energy consumption (J kg1 s1 ) due to standard metabolic rate, the
cost of gas secretion, and activity metabolism caused by compensatory swimming speed when the swimbladder volume deviated from optimal.
The standard metabolic rate (SMR) is identifiable as the baseline (swimbladder volume optimal and thus no need for compensatory swimming).
(d) The black bars indicate when the model predicted that hovering would have been an energetically favourable option.
4. Discussion
The presentation of this model is based on parameter
values for cod. In a strict sense, the results apply only to
cod. However, throughout this discussion of buoyancy
regulation we try to establish a wider perspective, aiming at making general points with validity for a broad
323
324
325
Acknowledgements
5. Conclusions
According to the resolution of the time scale, it is
possible to use the equations described here to study
behaviour in varying detail. Although the numerical
estimates of energy consumption are to some degree
uncertain, it will be possible to use the qualitative estimates to compare both behavioural strategies and geographical and temporal variations from longer timeseries. In a more detailed perspective, the energetics
during daily vertical migrations in combination with
We would like to thank K. Michalsen for kindly providing us with the depth versus time data set on which
we were able to visualise the model, A.G.V. Salvanes
for data on fish fins, and I. Huse for reading and commenting the manuscript. We would also like to thank
two anonymous reviewers for valuable comments. E.S.
was supported by The Research Council of Norway
and the European Commission contract Q5RS-200200813; C.J. and G.H. were supported by The Research
Council of Norway.
326
References
Alexander, R.M., 1959. The physical properties of the swimbladders
of fish other than Cypriniformes. J. Exp. Biol. 36, 347355.
Alexander, R.M., 1966. Physical aspects of swim bladder function.
Biol. Rev. Camb. Philos. Soc. 41, 131176.
Alexander, R.M., 1971. Swimbladder gas secretion and energy expenditure in vertically migrating fishes. In: Farquhar, G.B. (Ed.),
Proceedings of an International Symposium on Biological Sound
Scattering in the Ocean. Maury Center for Ocean Science, Washington DC, pp. 7485.
Alexander, R.M., 1972. The energetics of vertical migration by
fishes. Symp. Soc. Exp. Biol. 26, 273294.
Alexander, R.M., 1990. Size, speed and buoyancy adaptations in
aquatic animals. Am. Zool. 30, 189196.
Alexander, R.M., 2003. Principles of Animal Locomotion. Princeton
University Press, Princeton, NJ, 371 pp.
Arnold, G.P., Greer Walker, M., 1992. Vertical movements of cod
(Gadus morhua L.) in the open sea and the hydrostatic function
of the swimbladder. ICES J. Mar. Sci. 49, 357372.
Beamish, F.W.H., 1966. Vertical migration by demersal fish in the
Northwest Atlantic. J. Fish. Res. Board Can. 23, 109139.
Blake, R.W., 1979. Energetics of hovering in the mandarin fish (Synchropus picturatus). J. Exp. Biol. 82, 2533.
Brunel, P., 1965. Food as a factor or indicator of vertical migrations
of cod in the western gulf of St. Lawrence. ICNAF Spec. Publ.
6, 439448.
Clark, C.W., Levy, D.A., 1988. Diel vertical migration by juvenile
sockeye salmon and the antipredation window. Am. Nat. 131,
271290.
Clay, C.S., Horne, J.K., 1994. Acoustic models of fish: the Atlantic
cod (Gadus morhua). J. Acoust. Soc. Am. 96, 16611668.
Denton, E.J., Liddicoat, J.D., Taylor, D.W., 1972. The permeability
to gases of the swimbladder of the conger eel (Conger conger).
J. Mar. Biol. Assoc. U.K. 52, 727746.
Dickson, K.A., Donley, J.M., Sepulveda, C., Bhoopat, L., 2002. Effects of temperature on sustained swimming performance and
swimming kinematics of the chub mackerel Scomber japonicus.
J. Exp. Biol. 205, 969980.
Drucker, E.G., Lauder, G.V., 1999. Locomotor forces on a swimming fish: three-dimensional vortex wake dynamics quantified
using digital particle image velocimetry. J. Exp. Biol. 202, 2393
2412.
Fange, R., 1953. The mechanisms of gas transport in the euphysoclist
swimbladder. Acta. Physiol. Scand. 30 (Suppl. 110), 1133.
Glasstone, S., Lewis, D., 1964. The Elements of Physical Chemistry.
MacMillan & Co, London, 758 pp.
God, O.R., Michalsen, K., 2000. Migratory behaviour of north-east
Arctic cod, studied by use of data storage tags. Fish. Res. (Amst)
48, 127140.
Hansson, S., Rudstam, L.G., Kitchell, J.F., Hilden, M., Johnson, B.L.,
Peppard, P.E., 1996. Predation rates by North Sea cod (Gadus
morhua)predictions from models on gastric evacuation and
bioenergetics. ICES J. Mar. Sci. 53, 107114.
Harden Jones, F.R., 1951. The swimbladder and vertical movements of teleost fishes. I. Physical factors. J. Exp. Biol. 28, 553
566.
327